How the US and EU Are Building a Transatlantic AI Partnership

A one-page US draft proposal, circulated this week, is pushing for a formal AI partnership with the EU built around semiconductor supply chains and regulatory alignment, but the EU’s own digital sovereignty agenda is already pulling in a different direction. The EU formally joined the US-led Pax Silica framework around June 23, 2026, alongside 23 other nations, yet several member states remain openly sceptical. How the two blocs manage that tension will determine whether the partnership produces binding commitments or remains a statement of shared intent.

Phase 1: Laying the Foundational Frameworks for Cooperation

The first stage of building a transatlantic AI partnership involves identifying and working through existing multilateral and bilateral structures. These provide the platforms for dialogue, information exchange and gradual alignment on AI governance and supply chain security, areas where uncoordinated action carries real strategic risk.

The US-led Pax Silica framework, launched in December 2025, is the most visible of these structures. It is a non-binding initiative covering secure supply chains across critical minerals, energy inputs, advanced manufacturing, semiconductors and AI technology. When the EU, Germany, Greece and the Netherlands formally joined around June 23, 2026, total signatories reached 24 countries. The framework carries no dedicated funding or enforceable mandates at this stage, but its expanding membership reflects a degree of political convergence on the China supply chain question that would have been harder to achieve two years ago.

Alongside Pax Silica, the US-EU Trade and Technology Council (TTC), established in June 2021, remains the primary institutional channel for sustained bilateral coordination. The TTC has developed a Joint Roadmap for transatlantic AI cooperation and established early-warning systems for supply chain disruptions, work that does not require new legislation and can move faster than formal treaty-making. It also serves as the main forum for managing friction around the EU AI Act’s extraterritorial reach, a subject likely to dominate TTC discussions through 2026 and beyond.

Phase 2: Securing Critical AI Supply Chains

Supply chain security is where the partnership has its most concrete commitments so far. The EU’s pledge to purchase at least $40 billion worth of US AI chips is the clearest signal that geopolitical alignment is being backed by commercial ties, not just diplomatic language. It creates a direct demand signal for US semiconductor manufacturers and reduces the EU’s exposure to single-source dependency in a critical sector.

Export control coordination is the harder piece. The strategic logic behind Pax Silica explicitly includes curbing China’s access to advanced AI and semiconductor technology, which requires partner nations to align their export licensing regimes, a process that cuts against national commercial interests when domestic firms want to sell into the Chinese market. Co-investment in advanced chip manufacturing is the complementary mechanism: if allies collectively build more production capacity, they reduce the economic pressure to maintain access to adversarial markets.

Both the US and EU have made substantial legislative moves to build domestic semiconductor capacity. The US CHIPS Act, enacted in 2022, committed tens of billions in subsidies to domestic manufacturing. The EU’s Chips Act followed in 2023, facilitating public-private investment exceeding €31.5 billion for production facilities and pilot lines. By 2026, the EU had proposed a Chips Act 2.0 to accelerate further. These parallel efforts diversify global manufacturing capacity and reduce the concentration risk that currently sits heavily in Taiwan, a shared vulnerability neither bloc has fully resolved.

Phase 3: Navigating AI Regulatory Divergence

Regulatory alignment is where the partnership faces its most structural difficulty. The EU AI Act, the world’s first comprehensive AI law, applies extraterritorially, meaning US companies serving EU customers are subject to its provisions regardless of where they are incorporated. As of August 2, 2026, organisations become subject to the transparency obligations in Article 50, covering governance, disclosure and content-labelling requirements for certain AI systems. The high-risk system obligations under Annex III were provisionally delayed to December 2, 2027, while General Purpose AI model obligations have been enforceable since August 2, 2025.

The proposed Cloud and AI Development Act (CADA) adds another layer. It aims to establish a cloud and AI sovereignty framework based on assurance levels covering EU data location and independence from third-country providers, which, in practice, means US providers. For American companies, CADA would introduce compliance obligations that sit on top of, and in some respects conflict with, the operating model they have built for the EU market.

Against this backdrop, the Joint Statement on AI Opportunity signed by the US and nearly three dozen Pax Silica economies, including Germany, Greece and the Netherlands, offers a basis for common principles rather than harmonised rules. The statement centres on a pro-growth, pro-innovation approach and emphasises the role of the private sector. It does not resolve the gap between the EU’s risk-based, human-centric regulatory model and the US’s more industry-led approach, but it creates a reference point for future negotiation, potentially through regulatory sandboxes or shared best-practice frameworks. Whether that proves sufficient for companies operating under both regimes is a different question. For more on how AI agent deployments are being reassessed under rising compliance pressure, see our coverage of enterprise AI agent rollbacks.

Phase 4: Addressing Digital Sovereignty and Geopolitical Tensions

The EU’s push for digital sovereignty is not a negotiating position, it is a structural feature of European technology policy that any US partnership proposal has to account for. The European Commission’s “European Technological Sovereignty Package,” published on June 3, 2026, targets Europe’s declining share in key digital markets: EU providers currently hold around 15% of the cloud services market. The package includes CADA and Chips Act 2.0, and it is explicitly framed around reducing dependence on foreign, primarily US, providers in cloud, AI, semiconductors and open-source software. Treating this as mere protectionism misreads the political dynamics; it is a response to a genuine strategic vulnerability that European governments across the political spectrum have identified.

France’s position illustrates the internal divisions the US proposal must navigate. French officials have reportedly described Pax Silica as an attempt to colonise Europe’s technology sector, framing it as incompatible with the EU’s sovereignty agenda. Germany, Italy and the Netherlands have shown more support for the US approach. The divergence matters practically: EU technology policy requires broad member-state consensus, and a coalition of sceptics led by France can slow or reshape any agreed framework. The US will need to demonstrate that partnership terms offer genuine mutual benefit rather than preferred access to the European market under a security banner.

From the US side, the constraints are equally real. President Trump’s Executive Order on Promoting Advanced Artificial Intelligence Innovation and Security, signed on June 2, 2026, directs federal agencies to modernise government systems using AI capabilities, harden them against external threats and protect American AI intellectual property from exploitation and theft. That mandate will shape what the US is willing to share under any partnership agreement and where it draws limits on technology transfer. A workable transatlantic AI partnership ultimately has to operate within those constraints on both sides, which is why the current framework remains non-binding and the harder questions of regulatory harmonisation remain open. For broader context on how the US Congress is approaching AI governance alongside executive action, see our coverage of Pentagon AI acceleration and congressional oversight.

The US proposal and the EU’s Pax Silica accession represent genuine diplomatic momentum, but the structural divergences, on regulation, sovereignty and the terms of technology sharing, are real and are not resolved by a non-binding framework. The $40 billion chip purchase pledge gives the partnership commercial weight; the EU’s sovereignty package and French-led scepticism give it friction. Whether the four-stage process outlined here produces enforceable coordination or remains a set of parallel national strategies depends on whether both sides are willing to constrain their own regulatory autonomy in exchange for collective security. For more coverage of AI policy and regulation, visit our AI Policy & Regulation section.